Furbringer J-M
Bibliographic info:
15th AIVC Conference "The Role of Ventilation", Buxton, UK, 27-30 September 1994

Simulation is proving more and more important in building physics. Programs of different levels of complexity are today available for researchers and designers to model and plan buildings. But the accuracy of the output is not usually provided as a common result. This paper is a short summary of a dissertation [1] focused on the accuracy of the simulation outputs as a function of the accuracy of the input parameters. This is a point which requires particular attention, so that the simulation outputs can be used with their confidence intervals; without these intervals, the use of the simulation output is risky. The following question is discussed in the paper : is the prediction of detailed models more accurate than that of simple models if the accuracy of their respective input parameters is taken into account ? There is a risk that inaccurate input data can invalidate attempts at exact simulation. For the studied case, the answer is that the detailed model has larger confidence intervals than the simple models in wind as well as stack dominated situations. The result has been obtained by investigating models and measurement processes and determining their confidence intervals. Fractional factorial design has been used to estimate the partial derivatives of the models by the input parameters with an optimum number of simulations for the detailed multizone model COMIS and 4 simple models BREVENT, LBL, AIDA and TURBUL. A 6-zone family house was chosen as case of study because it allowed the comparison of the sensitivity of simple and detailed models.